The model doesn't execute functions — it only outputs structured call instructions. Execution is done by application code. Watch a complete two-step tool-calling flow: "Check weather → Conditionally send email."
1
Define Tools & Send Request
2
LLM Decision
3
Execute Tool
4
Second Decision
5
Generate Response
👤
User / App Code
🧠
LLM (GPT-4o)
⚡
Tool Engine
?
Ready
Click "Start Demo" to watch the Agent automatically complete a two-step tool-calling flow: "Check weather → Conditionally send email."